Oleaginous yeasts for biochemicals, biofuels and food from lignocellulose‐hydrolysate and crude glycerol

Microbial lipids produced from lignocellulose and crude glycerol (CG) can serve as sustainable alternatives to vegetable oils, whose production is, in many cases, accompanied by monocultures, land use changes or rain forest clearings. Our projects aim to understand the physiology of microbial lipid production by oleaginous yeasts, optimise the production and establish novel applications of microbial lipid compounds. We have established methods for fermentation and intracellular lipid quantification. Following the kinetics of lipid accumulation in different strains, we found high variability in lipid formation even between very closely related oleaginous yeast strains on both, wheat straw hydrolysate and CG. For example, on complete wheat straw hydrolysate, we saw that one Rhodotorula glutinis strain, when starting assimilating D‐xylosealso assimilated the accumulated lipids, while a Rhodotorula babjevae strain could accumulate lipids on D‐xylose. Two strains (Rhodotorula toruloides CBS 14 and R. glutinis CBS 3044) were found to be the best out of 27 tested to accumulate lipids on CG. Interestingly, the presence of hemicellulose hydrolysate stimulated glycerol assimilation in both strains. Apart from microbial oil, R. toruloides also produces carotenoids. The first attempts of extraction using the classical acetone‐based method showed that β‐carotene is the major carotenoid. However, there are indications that there are also substantial amounts of torulene and torularhodin, which have a very high potential as antioxidants.


| INTRODUCTION
The current economy heavily depends on fossil resources. For example, in fuel production, many branches of the chemical and food industry rely heavily on petrol polymers. This is problematic due to the finite nature of fossil resources and their alarmingly strong negative influence on climate change. The major challenge for humankind is to change to a bio-based economy, using renewable resources in a sustainable way. Biomass is seen as a potential raw material to replace fossil resources (Yang et al., 2021). Strictly speaking, fossil resources also derive from biomass origins but are highly reduced in oxygen and stored for a very long time (Sato, 1990).
When utilising fossil fuels, the stored carbon is released into the atmosphere as carbon dioxide, whereas new biomass binds carbon dioxide. Thus using new biomass will lead to a more balanced carbon cycle (Spagnuolo et al., 2019).
However, the replacement of fossil resources by biomass can result in other problems, such as competition for arable land, land-use changes in natural ecosystems, and in some cases even increased GHG release when compared to fossil-based systems (Gontard et al., 2018). These challenges can at least partially be approached by the use of byproducts, co-products and organic waste from agriculture and forestry, that is, nonedible lignocellulosic residues (Gontard et al., 2018;Valentine et al., 2012). Lignocellulose is rich in oxygen and also has the potential to generate other chemicals than those generated from substrates of petrochemical origin. However, for fuel production or in the oleochemical industry, there is a need to have a reduced oxygen content in the feedstock (Demirbas, 2011).
This review highlights some recent developments in research on oleaginous yeasts, their physiology on various carbon sources, the application of yeast lipids as feed and the production of potential other products besides lipids.

| PHYSIOLOGY OF LIPID ACCUMULATION IN OLEAGINOUS YEASTS
Oleaginous yeasts accumulate lipids as storage lipids, usually in the form of triacylglycerols (TAGs). However, also a substantial proportion of free fatty acids has been recently found in several oleaginous yeasts (Nagaraj et al., 2022;Shapaval et al., 2019). Lipid accumulation takes place when there is a surplus of a carbon source (Dglucose, etc.) combined with limited availability of nutrients, such as nitrogen, phosphorus, sulphur, and so forth (Granger et al., 1993).
Under these conditions, carbon flux is directed toward lipid synthesis.
Sugars are converted via glycolysis and pentose-phosphate pathway (PPP) to pyruvate, which is transported into the mitochondria and oxidatively decarboxylated to acetyl-CoA by pyruvate dehydrogenase. The acetyl-CoA is further metabolised via the tricarboxylic acid (TCA) cycle ( Figure 1). Pentoses are metabolised via the nonoxidative PPP. The most abundant pentose xylose is, as in most fungi, first converted to xylitol with an NADPH-dependent xylose reductase, which is then re-oxidised by an NAD + -dependent xylitol dehydrogenase to xylulose. Xylulose is converted by xylulokinase to the PPP-metabolite xylulose-5-P. However, xylulose-5-P can also be converted to glyceraldehyde-3-P and acetyl-CoA by a phosphoketolase reaction (Ratledge & Wynn, 2002). Recently, an alternative pathway of xylulose utilisation has been suggested in R. toruloides, where it is first reduced to D-arabitol, and then re-oxidised to ribulose, which can then be phosphorylated and metabolised via the PPP (Jagtap et al., 2021;Jagtap & Rao, 2018 On reaching a critical value, citrate is transported out of the mitochondria into the cytoplasm in an exchange with malate, presumably by a citrate/malate shuttle (Evans et al., 1983). Citrate is subsequently cleaved into acetyl-CoA and oxaloacetate by ATPcitrate lyase, a key enzyme present in the cytosol in oleaginous microorganisms. This is at the expense of ATP and the opposite reaction of citrate synthetase in the TCA cycle. The increasing amount of acetyl-CoA is further shuttled into fatty acid synthesis, which takes place in the cytoplasm. The formed oxaloacetate is

Take-away
• Methods for production and rapid quantification of yeast lipids established.
• Huge diversity in sugar assimilation and lipid formation in oleaginous yeasts.
• Co-production of chemicals to improve the feasibility of lignocellulose conversion.  Figure 1). Under phosphate limitation, AMP is dephosphorylated to adenosine and phosphate, to provide phosphate for cellular processes such as nucleic acid synthesis. This also results in the inactivation of ICDH, export of citrate from the mitochondria and thus acetyl-CoA accumulation in the cytoplasm. However, during P-limitation, a lower flux through the PPP has been observed, resulting in a limitation of NADPH, which is required as an electron donor for lipid synthesis (Wang et al., 2018).
Acetyl-CoA is converted into malonyl-CoA by acetyl-CoA carboxylase. In the next step, the fatty acid synthase (FAS) complex forms acyl-CoA from acetyl-CoA and malonyl-CoA, where the acyl-CoA chain is prolonged by two C-atoms, and a CO 2 is released. For Acyl-CoA formation, NADPH is required. This in most cases is probably generated by D-glucose-6-phosphate dehydrogenase and 6-phosphogluconate dehydrogenase in the PPP, or by a maleic enzyme (ME) (Ratledge, 2014). In some cases, for example, in L.
starkeyi, ME uses NAD + , not NADP + as a cofactor, and might, thus, not be involved in fatty acid synthesis (Tang et al., 2010). The acyl-CoA chains produced are transferred to the endoplasmic reticulum, where esterification with glycerol-3-phosphate (G-3-P) takes place to generate either structural lipids (such as phospholipids or glycolipids) or storage lipids in the form of TAGs (Fakas, 2017). Lipids are stored in intracellular lipid droplets (LD), also called lipid bodies. Their biogenesis starts between the two membrane leaflets of the endoplasmic reticulum. The mechanisms involved are not yet fully understood, but enzymes connected to the TAG synthesis have been found in these organelles. It has been proposed that LD grow by synthesising TAGs on their surface (Athenstaedt et al., 2006;Garay et al., 2014;Zanghellini et al., 2010).
The mechanism of fatty acid synthesis is conserved between eukaryotes and bacteria, the major difference is the structure of FAS. Some bacteria form branched-chain fatty acids (Cronan & Thomas, 2009). Many bacteria accumulate polyhydroxyalkanoates as storage molecules, but in some actinomycetes, cyanobacteria and Rhodococcus species also TAGs are accumulated (Dourou et al., 2018).

| LIPID PRODUCTION FROM LIGNOCELLULOSIC RAW MATERIALS
Lignocellulose mainly consists of three organic polymers cellulose, hemicellulose and lignin. The composition of lignocellulose can vary depending on the plant species, but also on the growth conditions of F I G U R E 1 Simplified scheme of lipid accumulation in oleaginous yeasts. Nitrogen and phosphate limitation results in the degradation of adenosine monophosphate, which is an essential cofactor of isocitrate dehydrogenase. Isocitrate is not further metabolised, accumulates and is converted to citrate, which is in equilibrium with isocitrate. Citrate is exported to cytosol with the citrate-malate shuttle. In the cytosol, citrate is degraded to acetyl-CoA and oxaloacetate. The latter is converted to malate and transported back to the mitochondria. Acetyl-CoA is the precursor of fatty acid (FA)-synthesis, which is finally converted to triacylglycerol (TAG) and stored in lipid droplets. Pentoses like xylose are converted to lipids via the pentose-phosphate pathway or a phosphoketolase reaction (see text, modified from Chmielarz [2021]).
the respective plant. Apart from organic polymers, lignocellulose also contains inorganic compounds, which remain as ash after burning.
The content varies, but for instance in rice straw, ash content can be up to 20%. The major element within the ash is Si, amorphous Si polymers have been observed to form incrustations in the epidermis, vascular bundle and other plant tissues. Si polymers may inhibit cellulase enzymes, case scaling or fouling of equipment (Satlewal et al., 2018).
D-Cellulose is a homo-polysaccharide composed of cellobiose subunits, that is, D-glucose linked by β−1,4-glycosidic bonds, packed tightly in cellulose fibres. It has a crystalline structure and is composed of several hundred to tens of thousands of glucose monomers (Nishiyama et al., 2003). Native cellulose has a high degree of polymerisation, is insoluble in water and is difficult to hydrolyse.
Hemicellulose is a hetero-polysaccharide, which structure can vary between plants, plant tissues and even within molecules ( Lignocellulose has evolved to provide rigidity to plant cell walls and a barrier against infection by microbes (Gupta et al., 2016;Maity, 2015). This implies that energy-intensive pretreatment and enzymatic degradation of lignocellulose are required, to make the carbon sources bound in it available for the yeast. A variety of pretreatment methods has been developed. Pretreatment is out of the scope of this article, there are several excellent recent reviews dedicated to this topic (Andlar et al., 2018;Galbe & Wallberg, 2019). Apart from sugars, also inhibitors including organic acids (which also can be carbon sources for oleaginous yeasts), sugar anhydrides such as furfural and hydroxymethylfurfural and aromatic compounds are released during pretreatment (Jönsson & Martín, 2016), which have an impact on subsequent substrate conversions. Pretreatment is actually a major driver of processing costs in lignocellulose-based processes. The costs of palm oil are currently estimated at $901.50/t (https://www.indexmundi.com/commodities/?commodity=palm-oil% 26months=60). Estimates of the minimal selling prices of lignocellulose-derived yeast oil range from $1800 to more than $10,000 per metric ton. To obtain a cost-competitive process, a biorefinery approach, valorising co-products of oil production is crucial (Parsons et al., 2020).
Combined production of biodiesel and biogas from L. starkeyi biomass grown on wheat straw hydrolysate, as well as using the residues for the production of electricity and heat was shown to result in a fossil fuel replacement potential of 5.74 MJ/kg straw. Utilising the lignin to obtain process energy and not returning it to the soil had the highest impact on greenhouse gas savings. Improved lipid extraction and shortening the energy-intensive aerobic fermentation process were identified as the major crucial factors for improving the energy output and greenhouse gas balance of biodiesel production based on microbial lipids. Utilisation of crude glycerol (CG) for biodiesel production was also an important factor (Karlsson et al., 2016(Karlsson et al., , 2017. Inhibitor tolerance would be one important factor to decrease fermentation times. A variety of detoxification procedures for lignocellulose hydrolysate has been developed, however, those steps should be minimised, as detoxification can represent up to 22% of second-generation production costs (Koppram et al., 2012). Substrate toxicity can also be overcome by isolating inhibitor-resistant mutants or adapting strains to high concentrations of inhibitors . Obtaining stable mutants and further development of genetic tools may identify genetic and metabolic networks crucial for inhibitor tolerance.
Diluting a substrate would also be an option for reducing the inhibitor concentrations, however, this would also dilute the carbon sources for lipid production. This can be avoided by running cultures in a fed-batch mode, which gives the cells the opportunity to adapt to increasing inhibitor concentrations. Those strategies have frequently been used for converting hemicellulose hydrolysate to lipids.
Hemicellulose hydrolysates are frequently generated as a side product of the pulp-and paper industry. They can be used for steam and electricity production, but the heating value is relatively low (Helmerius et al., 2010;H.-J. Huang et al., 2010). Hemicellulose hydrolysate is also generated in acid pretreatment of lignocellulosic biomass, which is a quite common method for pretreatment (Galbe & Wallberg, 2019). reported above for birch wood hydrolysate (Xavier et al., 2017).
As mentioned above, pentoses from hemicellulose can be converted to their anhydride furfural, which is a fermentation inhibitor. On the other hand, furfural is classified as an important platform chemical, which is exclusively produced from plant biomass.
Established methods for generating furfural from lignocelluloses damage the other polysaccharides in a way that it is almost impossible to further convert them by microbial conversion.
However, a few years ago a novel catalytic method has been developed to produce furfural, leaving the cellulose fraction intact (Vedernikovs, 2020). The combined production of furfural and ethanol or lipids from wheat straw has been investigated. The cellulose hydrolysate had a relatively low inhibitory potential, no dilution or detoxification was required. Ethanol production resulted in a theoretical yield from the released sugar. In total, a production potential of 110 g furfural and 111 g ethanol or 33 g lipids from 1 kg of wheat straw was determined. Lipid production was tested with L.
starkeyi CBS 1807 and R. babjevae DBVPG 8058. Interestingly, the two strains showed strong differences in their lipid production potential. L. starkeyi reached a lipid yield of 0.09 g per g consumed D-glucose, and a production rate of 0.08 g/l h; R. babjevae 0.17 g/g and 0.18 g/l h . Similarly, furfural could also be produced from rapeseed oil, and ethanol production from the cellulose fraction was demonstrated (Rozenfelde et al., 2021).
Another opportunity for reducing the concentration of inhibitors would be mixing them with another low-value carbon source, for instance, CG. CG is a residue from the production of biodiesel, where the glycerol in the triglycerides is replaced by methanol, forming fatty acid methyl esters. CG contains several toxic components like soaps, ash and methanol, which can be removed by for instance distillation, probably also other mechanisms responsible for the metabolic activation, for example, the availability of additional nitrogen sources in the hydrolysate or the activation of some stress response, which then might result in a higher substrate uptake (Chmielarz, 2021).

| METHODS FOR LIPID QUANTIFICATION
Probably the most commonly used methods for lipid quantification in oleaginous yeasts are based on the Folch or Bligh and Dyer extraction methods (Bligh & Dyer, 1959;Folch et al., 1957). Both methods have been designed for animal tissues and are based on extraction with organic solvents and final gravimetric lipid quantification. For microorganisms such as yeasts, the methods have to be adapted due to the complex and rigid cell wall and membrane structures (Jacob, 1992). Various mechanical, chemical or enzymatic pretreatments of cell biomass have been tested to disrupt the cells and make most of the intracellular lipids accessible. This includes, for example, acid-catalysed hot water treatment, autoclaving, bead beating, homogenisation, microwave radiation, ultrasonication or thermolysis (Bonturi et al., 2017;Brandenburg et al., 2018;Patel et al., 2019 (Kimura et al., 2004;Sitepu et al., 2012). On the other hand, this staining gives impressive results while using fluorescent microscopy for visualisation of intracellular lipid bodies and could be used in screening studies (Shi et al., 2015).
Spectroscopy is a nondestructive method that allows rapid and noninvasive investigation of biomolecules, for instance, proteins or lipids. Near infrared (NIR) or mid infrared (MIR) light is close to the visible light spectra, with wavenumbers between 12,500-4000 and 4000-400 cm −1 , respectively. MIR spectroscopy, commonly called IR spectroscopy, can be used to study the fundamental vibrations of molecules. With the NIR spectroscopy overtones and combination bands are detected. The advantage of MIR is that the absorption spectrum provides more clear peaks referring specifically to functional groups. With NIR the spectra are broader and more undefined but may hold more information. Fourier transform (FT)-NIR applies additional equipment, interferogram, which allows plotting light intensity as a function of time, which is then Fourier transformed to a frequency domain. FT-(N)IR is a much faster technique allowing to measure the whole wavelength range simultaneously, is more sensitive and has less PASSOTH ET AL.
The advantages of this type of spectroscopy are that almost no sample preparation is required, besides simple washing and drying, and results are given almost in real-time. It reduces costs, labour, analysis time and a lot of solvent usage. In addition, the amount of needed sample volume can be decreased tremendously, making it possible to work in small-scale experiments for screening or lipid kinetics studies (Ami et al., 2014;Brandenburg et al., 2021;Chmielarz et al., 2021;Forfang et al., 2017;Kosa et al., 2017).
Quantitative measurements with FT-(N)IR must always be calibrated against a set of reference values, determined by classical lipid extraction and analysis, creating a prediction model. As water molecule vibrations cause huge responses in FT-IR and -NIR spectroscopy, it strongly interferes with the lipid prediction models.
Therefore, it is important to dry the cells before measurement.
Prediction models can also be influenced by the yeast strain used in measurements. FT-NIR models for the prediction of total lipids were established for L. starkeyi and R. toruloides. The models had a high accuracy with R 2 -values of 96% and 98%, respectively. The L. starkeyi model enabled predicting the lipid content in L. starkeyi and Y.
lipolytica grown under several conditions, but not that of strains of pigmented yeasts, that is, Rhodotorula spec., similarly, the Rhodotorula model was not appropriate to predict nonpigmented strains. A combined model was established by including strains of L. starkeyi, Y. lipolytica, R. toruloides. For testing the models, also strains of R. glutinis and R. babjevae were included. This model enabled the prediction of the lipid content in all yeast strains but with lower accuracy than the specific models ( Figure 3).
In another approach, an FT-IR method was established, based on 13 oleaginous yeast strains. Different spectral regions and several spectral processing methods were investigated to build up robust and accurate prediction models for total lipid amount and lipid profile. Spectral region evaluation showed that the spectral range for 3100-2800 cm −1 was most reliable for the prediction of total lipids, saturated, monounsaturated and polyunsaturated fatty acids; ═C─H stretching; C─H asymmetric stretching of CH 3 ; asymmetric stretching >CH 2 of acyl chains; symmetric stretching of CH 2 of acyl chains. In addition, the spectral region 1800-700 cm −1 was also used for total lipid determination, including the peak for C═O stretching and regions that refer to proteins, CH 2 deforming, CH 3 bending, C─O─C stretching, carbohydrates and polyphosphates ( Figure 4). For some strains, a peak at 1710 indicates the presence of free fatty acids. This was confirmed by thinlayer chromatography (Shapaval et al., 2019). In some strains, free fatty acids were the second most abundant class of lipids, which was also confirmed in a recent study in R. toruloides CBS 14 (Nagaraj et al., 2022).

| STRAIN DIFFERENCES IN THE ABILITY TO ACCUMULATE LIPIDS
Using the above-described spectroscopic methods, a variety of oleaginous strains were tested on D-glucose, D-xylose, CG and lignocellulose hydrolysate. A huge physiological diversity even between strains belonging to the same or closely related species was observed (Brandenburg et al., 2021;Chmielarz et al., 2021).
Among 13 tested strains of Lipomyces spp., four were not able to grow on D-xylose (Brandenburg et al., 2021), although these species are usually characterised by a good ability to grow on this substrate (Kurtzman et al., 2011;Sitepu et al., 2014)  CG has been tested as a substrate for oleaginous yeasts in a variety of studies (e.g., Chatzifragkou et al., 2011;Diamantopoulou et al., 2020;Papanikolaou & Aggelis, 2009;Posada & Cardona, 2010). Chmielarz et al. (2021) tested 27 strains for growth on this substrate.
Interestingly, only 11 of them were able to grow on the tested CG (obtained from Perstorp AB, Sweden), using FT-NIR for following lipid accumulation over time. Out of the 11 tested Lipomyces-strains, only L. starkeyi CBS 7786 was able to grow on CG. However, it was growing slower than the identified positive Rhodotorula strains and was strongly inhibited at CG concentrations of 60 g/L. Out of the 16 tested Rhodotorula strains, 10 were able to grow on CG. All of these strains were able to grow up to CG concentrations of 60 g/L, although with different growth rates. As mentioned above, mixing the CG with hemicellulose hydrolysate had a positive effect on growth and lipid accumulation by Rhodotorula strains, but not by L. starkeyi.
The above-mentioned results show that there is a huge diversity in lipid accumulation between strains even of closely related species.
Understanding the physiological basis of these differences would enable the identification of crucial steps in lipid production and thus F I G U R E 4 FT-IR spectrum of Rhodotorula toruloides grown on yeast extract peptone dextrose medium (P), nitrogen-limited medium (NLM) containing D-glucose (G), NLM containing D-xylose (X), NLM containing D-glucose and D-xylose (M

| CONCLUSIONS AND OUTLOOK
Oleaginous yeasts have great potential to replace vegetable oils in a variety of applications. Yeast oil has a similar composition to plant oil.
Apart from producing biodiesel from it, it can also be used as a feed additive Brunel et al., 2022), or even as food (Lundin, 1950). Carotenoids are formed by red yeasts and have great potential to be used in food and feed, and high-value chemicals . Especially the yeast-specific carotenoids torulene and torularhodin are of interest because they have a higher antioxidant potential than the most commonly used β-carotene (Ungureanu & Ferdes, 2012). There are reports about exopolysaccharides and glycolipids formed by Rhodotorula strains (Byrtusová et al., 2021). We also found exopolysaccharides and polyol esters of hydroxy-fatty acids formation in several strains (unpublished results).
These findings indicate a great biotechnological potential especially in the basidiomycetous red yeasts, also because they showed more rapid growth and lipid formation compared to the ascomycetous strains. A variety of efforts have been taken to improve the potential of oleaginous yeasts for lipid production, including optimisation of culture conditions, engineering metabolic pathways involved in producing building blocks for lipid synthesis and TAG assembly, F I G U R E 5 Growth and lipid content of Lipomyces and Rhodotorula strains grown in medium containing D-glucose (G), D-xylose (X) or a D-glucose/D-xylose mixture (M). The cultivations were done in shake flasks at 25°C, 130 rpm and performed in duplicates. Dark green represents strong growth and high lipid content, light green/white little or no growth and low lipid content. n.d., not determined due to no or very poor growth. Reprinted from Brandenburg et al. (2021).